To circumvent the “storage wall” and “power consumption wall” limitations inherent in traditional visual information processing systems, this study develops an ultra-low power two-terminal photoelectric synaptic device leveraging the pronounced persistent photoconductive effect of two-dimensional ReS₂. Employing a combination of first-principles calculations and experimental characterizations, we elucidate the regulatory mechanism of sulfur vacancies on the electronic density distribution and band structure of ReS₂. The introduction of sulfur vacancies induces defect energy levels within the band gap, elevates the local density of states, and promotes the separation and trapping of photogenerated electron-hole pairs. These mechanisms significantly amplify the persistent photoconductive effect, establishing a robust physical foundation for synaptic weight implementation. Notably, the device achieves an ultra-low energy consumption of 49 fJ per synaptic event, comparable to the energy efficiency of biological synapses. The synaptic weights can be continuously and controllably modulated by varying the intensity, number, and timing of optical pulses, accompanied by typical frequency-dependent plasticity. Leveraging its high-pass filtering characteristics, the device demonstrates effective edge enhancement in image preprocessing. Furthermore, by exploiting wavelength-dependent photo responses, the device successfully emulates the “Pavlovian dog” conditioned reflex, validating its capability for associative learning. This work unveils the sulfur vacancy-mediated photoelectric synaptic mechanism in ReS₂ at the atomic and electronic structure levels. It offers novel insights into balancing structural intricacy with ultra-low power performance, holding significant implications for the advancement of high-performance neuromorphic vision systems in edge computing.